Agricultural Productivit,y Economic Growth, and Food Security

BackgroundImpact assessment

The modelConclusions

Agricultural Productivity,

Economic Growth, and Food Security

Doug Gollin

Williams College

Agriculture for Development � Revisited

UC Berkeley, October 2010

Doug Gollin Agriculture and Growth



Outline

1 Background

2 Impact assessment

3 The model

4 Conclusions




Outline

1 Background

2 Impact assessment

3 The model

4 Conclusions




1. Conceptualizing agriculture in general equilibrium

Agriculture is a large sector in most poor countries, both in

terms of employment and output.

The agricultural sector has strong linkages to other sectors:

As a source of supply for a unique consumption goodAs a source of demand for non-agricultural productsAs a potential source of labor and other productive resources(land, capital)

These linkages make it important to think about agriculture in

general equilibrium.




Agricultural productivity⇒Macroeconomic outcomesMany channels of connection

Agricultural performance a�ects GDP levels and growth rates

Agricultural production often has big impacts on in�ation rates

Governments may earn large shares of revenue from

agricultural exports (e.g., export taxes in many African

countries)

Poverty alleviation and inequality




Macroeconomy ⇒Agriculture

Demand for products

Overall demand for labor

Exchange rates

Macro and monetary stability




Outline

1 Background

2 Impact assessment

3 The model

4 Conclusions




2. Impact assessment for agricultural interventions

Many agricultural interventions are targeted to the level of

farms or communities

New crop varietiesInput subsidiesRural roads and infrastructureEtc.




Bene�ts spill overAre we looking in the right place?

Evaluations often focus on the impact of these interventions

on farm income or living standards.

These interventions may create some bene�ts at the farm level.

But bene�ts may instead be transmitted through markets to

others.

Consumers may bene�t through lower prices.Processors may bene�t through more consistent supply orlower costs.Firms in other sectors may demand more labor and drive upwage rates.

Micro evaluations that do not consider general equilibrium

e�ects may give misleading results.




Example: Agricultural development

and structural transformation

Based on Gollin and Rogerson (2010), in a paper on Uganda.

Consider general equilibrium e�ects of programs and policies

that generate agricultural �development�.

These interventions cause the share of agriculture in GDP and

employment to fall.

Naive evaluation techniques might fail to recognize

agriculture's role.




Outline

1 Background

2 Impact assessment

3 The model

4 Conclusions




Related literature

Close in spirit to Eswaran and Kotwal (1993).

Some similarities to Caselli and Coleman (2001), Vollrath

(2004, 2008).

Also related to work on transportation and growth in

Herrendorf, Schmitz, and Teixeira (2006, 2008); Adamopoulos

(2005).

Most similar to models of structural transformation in Gollin,

Parente, and Rogerson (2004, 2007).




Model environment

Consider a closed developing economy with two sectors: an

agricultural sector that produces food and an urban sector that

produces non-agricultural goods.

Non-agricultural goods can be consumed or used as inputs into

the production process in either sector.

Non-agricultural goods are produced in the city.

Food can be produced in either of two rural regions: an area

�close� to the city or an area that is more �remote�.

There are (high) transportation and transaction costs that

make it expensive to move manufactured goods from the city

to rural areas and (symmetrically) make it costly to move food

from rural areas to the city.




Schematic Representation

Region 2 Region 0

a a

m m

Cost 1q Cost 2qRegion 1

City Manufacturing

Near Agriculture

Remote Agriculture




Model environment, cont.

In equilibrium, people will inhabit all three regions.

Those in the remote rural area will produce less for the market

than those in the close rural area.

The urban population is limited by the ability of the

agricultural sector to produce �marketable surplus;� low

agricultural productivity implies small urban populations.

High transportation costs will make it costly to move goods

across regions.




Three-Region ModelPreferences

Log linear preferences with non-homotheticities:

α log(a − a) + (1− α) log(m + m)




Three-Region ModelTechnologies

Agricultural technology:

aj = AaF (lj , xj , naj) = Aaj l1−θx−θnj xθxj nθnj

Manufacturing technology: m = Amnm.




Three-Region ModelEndowments

l1 = 0.1

l2 = 0.9

Labor is allocated endogenously.




Feasibility Conditions

n0m0 + n1m1 + x1

1− q1+ n2

m2 + x2

(1− q1)(1− q2)= Amn0

n0a0

(1− q1)+n1a1+n2(1−q2)a2 = Aal

1−θx−θn1 xθx1 nθn1 +(1−q2)Aal

1−θx−θn2 xθx2 nθn2




Equilibrium Allocations

For interior solution:

m0 + m =m1 + m

(1− q1)=

m2 + m

(1− q1)(1− q2)

a0 − a

(1− q1)= a1 − a = (1− q2)(a2 − a)

Corner solutions are plausible under some speci�cations. We

solve for them and check for them computationally.




Quantitative ExperimentParameter Values

Chosse parameter values to match a few stylized observations

from Uganda.

Aa = Am = 1θx = .2, θn = .4α = .20m = 0a⇒ n1 + n2 = 0.80q1 = 0.1, q2 = 0.6




Benchmark Allocations

Consumption Allocations: Three Region Model

n1 n2 a0 a1 a2 m0 m1 m2

.096 .707 .409 .410 .425 .0516 .0464 .0186

Individuals in the �near� agricultural region consume bundles

quite similar to urban residents.

Sell almost half (45%) of agricultural output.

Individuals in the �remote� agricultural region seem to be in

quasi-subsistence:

Consume only one-third as much m as urban residents.Sell only one third (34%) of own agricultural output.




Agriculture, Near and FarInputs and Intensity across Regions

Agriculture Production: Three Region Model

l1/n1 l2/n2 x1/n1 x2/n2 ya1/n1 ya2/n2 ya1/l1 ya2/l2

1.04 1.27 .187 .066 .73 .64 .70 .50

Labor intensity is greater in near region.

Far greater use of intermediates in near region.

Output per unit of land (yield) is 40% higher in the near

region.

Di�erences in output per worker are not large.




Comparative Statics

Consider three scenarios:

10% increase in agricultural TFP10% increase in manufacturing TFP10% reduction in transport cost

Welfare comparison: Ask how by what fraction the benchmark

consumption bundle would need to be increased in order to

yield the same utility as each scenario.




Equilibrium Scenarios

Experiments in the Three Region Model: Consumption Allocations

n0 n1 n2 a0 a1 a2 ∆

Benchmark .197 .096 .707 .409 .410 .425 −Aa = 1.1 .260 .115 .625 .415 .417 .442 .32

Am = 1.1 .210 .105 .685 .411 .412 .429 .06

q = .9q .259 .098 .643 .413 .414 .431 .26

Aa, Am, q .340 .124 .536 .420 .422 .448 1.07

Aa , q .320 .114 .566 .420 .421 .447 .82

l1 = .2 .280 .216 .504 .414 .415 .438 .35

Pop 1.1 .089 .099 .812 .407 .407 .418 −.02




Manufacturing Consumption in Equilibrium

Experiments in the Three Region Model: Manufacturing Consumption

m0 m1 m2 ∆

Benchmark .052 .046 .019 −Aa = 1.1 .096 .087 .035 .32

Am = 1.1 .065 .059 .024 .06

q = .9q .085 .077 .036 .26

Aa, Am, q .16 .15 .068 1.07

Aa , q .14 .13 .057 .82

l1 = .2 .095 .085 .034 .35

Pop 1.1 .036 .032 .013 −.02




Agriculture in EquilibriumIntensi�cation and Input Use

Experiments in the Three Region Model: Agricultural Production

ya1/n1 ya2/n2 ya1/l1 ya2/l2

Benchmark .73 .64 .70 .50Aa = 1.1 .75 .77 .86 .53Am = 1.1 .71 .66 .75 .51q = .9q .75 .72 .73 .51Aa, Am q .77 .91 .95 .54Aa,q .77 .86 .88 .54l1 = .2 .72 .75 .78 .47Pop 1.1 .71 .59 .70 .53




Agriculture in EquilibriumIntensi�cation and Input Use

Experiments in the Three Region Model: Agricultural Production

l1/n1 l2/n2 x1/n1 x2/n2

Benchmark 1.04 1.27 .187 .066Aa = 1.1 .87 1.44 .196 .080Am = 1.1 .95 1.31 .200 .075q = .9q 1.02 1.40 .221 .097Aa, Am q .81 1.68 .254 .137Aa,q .88 1.59 .225 .115l1 = .2 .92 1.59 .226 .094Pop 1.1 1.02 1.11 .176 .058




Impact assessmentThe perils of overlooking GE e�ects

In all of these scenarios, 80-90 percent of welfare gains come

from sectoral reallocation of people, in the sense that

increasing consumption allocations alone would generate only

10-20% of the welfare gains.

Impact assessment techniquest that focus on �within� measures

of welfare gains will give misleading (low) estimates of impact.

Similar in spirit to Lewis model, in which growth consists of

moving people from a low productivity sector to a higher

productivity sector.




Agricultural TFP increaseThe perils of overlooking GE e�ects

Consider the scenario in which agricultural TFP rises by 10%.

Our welfare measure suggests that this generates as much

improvement in well-being as a 32% increase in all

consumption allocations.

But suppose we do what is common in the agricultural

economics literature on impact assessment studies: we will

calculate increases in crop yields.




Sectoral impacts in agriculture

At a national level, yield increases by 7.7% (for a

population-weighted sample) or 8.3% (for an area-weighted

sample). Both are smaller in percentage terms than the TFP

increases.

Agricultural output increases by 8.6%.

We do not calculate prices, but shadow prices fall in the model

economy. The value of agricultural output rises only very

slightly.

But the big e�ect is that 6.3% of the people move from rural

to urban areas, while another 1.9% move from

quasi-subsistence to commercial agriculture.




Rural transportation infrastructure improvementsThe perils of overlooking GE e�ects

Consider the scenario in which transportation costs fall by

10%.

Our welfare measure suggests that this generates as much

improvement in well-being as a 35% increase in all

consumption allocations.

Suppose this policy is examined for its impact on rural

households.




Impacts on rural households

At a national level, yield increases by 1.4% (for a

population-weighted sample) or 2.3% (for an area-weighted

sample).

lAgricultural output increases by 0.44%;

Prices received by farmers may rise, but if we use market prices

for food, the decline in transport cost is likely to lead to

signi�cant decreases in market prices. Rural incomes rise very

slightly or perhaps fall.

Rural consumption rises � but nominal prices have fallen, so an

expenditure survey may not capture this adequately.

The big e�ect is that 20% of the total population is able to

move out of subsistence agriculture.




Implications

Need a general equilibrium model to think properly about

these impacts.

In evaluating economy-wide changes (e.g., Green Revolution,

any kind of national-level interventions), ignoring the GE

e�ects may lead to serious errors.

Surveys that track migration and sectoral movements will do a

better job than studies that focus only on rural or agricultural

households.




Attribution

Need to be careful with attribution of bene�ts:

Big gains occur through rural-to-urban migration.This movement may be driven powerfully by investments inagriculture.Empirical �nding that rural-to-urban migrants achieve thebiggest welfare gains will tell us nothing about the causalmechanisms; possibly this reallocation is best achieved throughinvestments in agriculture.




Outline

1 Background

2 Impact assessment

3 The model

4 Conclusions




5. Conclusions and Questions

We need to remember that in many developing countries,

agricultural interventions have impacts that spill out from the

sector. Will a macro perspective change our assessments of

which interventions are successful?

What government investments and policies can help to reduce

the costs of change?




Taking okro to market


Documents

Agricultural Productivit,y Economic Growth, and Food Security